A new Exponentially Expanded Robust Random Vector Functional Link Network based MPPT model for Local Energy Management of PV-Battery Energy Storage Integrated Microgrid. (May 2020)

• Record Type: Journal Article
• Authors: Priyadarshini, Lipsa; Dash, P.K.; Dhar, Snehamoy
• Is Part Of: Engineering applications of artificial intelligence. Volume 91(2020)
• Publication Date: 2020-05
• Subjects: Distributed adaptive droop -- Exponentially Expanded Robust Random Vector functional Link Network -- Local Energy Management -- Maximum power point tracking -- Photovoltaic systems; Engineering -- Data processing -- Periodicals; Artificial intelligence -- Periodicals; Expert systems (Computer science) -- Periodicals; Ingénierie -- Informatique -- Périodiques; Intelligence artificielle -- Périodiques; Systèmes experts (Informatique) -- Périodiques; Artificial intelligence; Engineering -- Data processing; Expert systems (Computer science); Periodicals; 620.00285
• DOI: 10.1016/j.engappai.2020.103633
• Languages: English
• ISSNs: 0952-1976

Abstract

Abstract: In this paper a new Maximum Power Point Tracking (MPPT) model is presented for Local Energy Management (LEM) of a multiple Photovoltaic (PV) based microgrid. To detect accurate MPP references under local uncertainties, a non-iterative Linear Recurrence Relationship (LRR) based PV model is incorporated with PV penetration index. A robust, accurate and fast Exponentially Expanded Robust Random Vector Functional Link network (EE-RRVFLN) based MPPT algorithm is constructed with an exponentially expansion unit to address positive dynamic volatility and a direct link relationship to address null vs. positive volatility in PV data. The robustness is further incorporated by a maximum likelihood estimator using Huber's cost function, where both input and output weights are optimally estimated by targeting reduction in MPP tracking error. An Assessment Index (i.e. MPPT error related) based Distributed Adaptive Droop (DAD) mechanism is suggested as Primary Controller (PC) for effective power sharing among multiple PVs. A detailed case study is presented to evaluate the accuracy of the proposed model in MATLAB simulation, as well as in dSPACE 1104 based Hardware-in-Loop (HIL) platform. Historical data for different intervals/ seasons, partial shading, improved LEM validations (simulation and HIL) are considered as different cases to establish the excellence of the proposed approach, as compared with conventional Functional Link Neural Network (FLNN) and Random Vector Functional Link Neural Network (RVFLNN). Graphical abstract: Highlights: A novel Exponentially expanded robust RVFLN is proposed for PV MPP Tracking. Local Energy Management task is accomplished using PV-Energy storage based microgrid. The RES-to-PV estimation is incorporated with a non-iterative PV equivalent model. Efficacy of proposed RVFLN is verified using historical and partial shading data. dSPACE based Hardware-in-Loop platform is used to validate performance of RVFLN-LEMS.